Intel Tiger Lake Xe Gen12 iGPU in unnamed Lenovo laptop puts up promising show against NVIDIA GeForce MX350 in Geekbench OpenCL test

[Redaktion]

A Geekbench 4 OpenCL benchmark surfaced last week pitting the Tiger Lake Core i7-1165G7 against the NVIDIA GeForce MX350. The Gen12 Xe iGPU in Tiger Lake manages to not only put up a convincing performance but also beats the MX350 in the test. We also get to see that the 80 EU Core i5-1135G7 is about 13.5% slower than a Core i7-1165G7 in the same benchmark.

https://www.notebookcheck.net/Intel-Tiger-Lake-Xe-Gen12-iGPU-in-unnamed-Lenovo-laptop-puts-up-promising-show-against-NVIDIA-GeForce-MX350-in-Geekbench-OpenCL-test.479608.0.html

[opelit]

The Xe iGPU have 50% more shaders than Vega8 in 4800U.
8x64 = 512 shaders
While 96EU in Intel is 96x8 = 768 shaders. So it's not surprise it perform better, the surprise will be if the performance will hold over time.

Overall
Vega8 in 4800U is 1,79Tflop (512 shaders at 1750mhz)
Xe in the 11gen of Intel will be probably 1,99Tflop (768 shaders at 1300mhz)

[Spunjji]

The Xe iGPU have 50% more shaders than Vega8 in 4800U.
8x64 = 512 shaders
While 96EU in Intel is 96x8 = 768 shaders. So it's not surprise it perform better, the surprise will be if the performance will hold over time.

Overall
Vega8 in 4800U is 1,79Tflop (512 shaders at 1750mhz)
Xe in the 11gen of Intel will be probably 1,99Tflop (768 shaders at 1300mhz)

They're certainly going the right way about enabling sustained performance - wider and slower tends to be a good approach. AMD seem to have boxed themselves in a little in terms of graphics performance with the decision to push for 8 CPU cores in Renoir.

[IntelUser2000]

They're certainly going the right way about enabling sustained performance - wider and slower tends to be a good approach. AMD seem to have boxed themselves in a little in terms of graphics performance with the decision to push for 8 CPU cores in Renoir.

This is often said, but it doesn't matter. AMD has said increasing clock speeds and reducing CU count ended up better with the 7nm process it was using.

If you reduce clocks you may be able to reduce voltage a bit, but not as much as you think. There's a reason they said not only clock scaling is dead, but so is voltage scaling.

[Tov]

This is a good news. Hope it can force AMD to put RDNA2 in their next APU instead of the Vega rumor.